Multimodal Causal Reasoning Benchmark: Challenging Vision Large Language Models to Discern Causal Links Across Modalities

TL;DR

This paper introduces MuCR, a benchmark for testing multimodal causal reasoning in large language models, revealing current limitations and proposing a strategy to improve their understanding of visual cues across modalities.

Contribution

The paper presents MuCR, a new benchmark for multimodal causal reasoning, and introduces VcCoT, a strategy to enhance visual cue recognition in large language models.

Findings

Current MLLMs underperform in multimodal causal reasoning compared to textual tasks.

Identifying visual cues is crucial for cross-modal generalization.

VcCoT improves the models' ability to leverage visual information for causal inference.

Abstract

Multimodal Large Language Models (MLLMs) have showcased exceptional Chain-of-Thought (CoT) reasoning ability in complex textual inference tasks including causal reasoning. However, will these causalities remain straightforward when crucial hints hide in visual details? If not, what factors might influence cross-modal generalization? Whether we can effectively enhance their capacity for robust causal inference across both text and vision? Motivated by these, we introduce MuCR - a novel Multimodal Causal Reasoning benchmark that leverages synthetic siamese images and text pairs to challenge MLLMs. Additionally, we develop tailored metrics from multiple perspectives, including image-level match, phrase-level understanding, and sentence-level explanation, to comprehensively assess MLLMs' comprehension abilities. Our experiments reveal that current MLLMs fall short in multimodal causal reasoning compared to their performance in purely textual settings. Additionally, we find that identifying visual cues across images is key to effective cross-modal generalization. Finally, we propose a VcCoT strategy that better highlights visual cues, and our results confirm its efficacy in enhancing multimodal causal reasoning. The project is available at: https://github.com/Zhiyuan-Li-John/MuCR